Deep targeted sequencing of circulating tumor DNA to inform treatment in patients with metastatic castration-resistant prostate cancer

Abstract

Background: Intrinsic and acquired resistance to second-generation anti-androgens pose a significant clinical challenge in the treatment of metastatic castration-resistant prostate cancer (mCRPC). Novel biomarkers to predict treatment response and inform alternative treatment options are urgently needed.

Methods: Deep targeted sequencing, with a prostate cancer-specific gene panel, was performed on circulating tumor DNA (ctDNA) and germline DNA from blood of mCRPC patients recruited in Denmark (n = 53), prior to starting first-line treatment with enzalutamide or abiraterone acetate, and for a subset of patients also at progression (n = 18). Likely clonal hematopoietic variants were filtered out. Genomic findings were correlated to clinical outcomes (PSA progression-free survival (PFS), overall survival (OS)). Intrinsic resistance candidate biomarkers were considered by enrichment analysis of nonresponders vs. responders. Genomic alterations at progression were considered as possible drivers of acquired resistance. Clinical actionability was assessed based on OncoKB and ESCAT.

Results: Somatic alterations in PTEN, cell cycle regulators (CCND1, CDKN1B, CDKN2A, and RB1) and chromatin modulators (CHD1, ARID1A) were associated with significantly shorter PFS and OS, also after adjusting for ctDNA% in multivariate Cox regression analysis. The associations with poorer outcomes for alterations in PTEN and chromatin modulators were validated in an external dataset. Patients with primary resistance to enzalutamide/abiraterone had enrichment for BRAF amplification and CHD1 loss, while responders had enrichment for TMPRSS2 fusions. AR resistance mutations emerged in 22% of patients at progression. These were mutually exclusive with other alterations that may confer resistance (i.e., activating CTNNB1 mutations, combined TP53/RB1 loss). Clinically actionable alterations, primarily in homologous recombination repair genes, were found in 54.7% and 49.0% of patients (OncoKB and ESCAT, respectively), with few additional alterations detected at progression. Level I alterations were identified in 41.5% of patients employing OncoKB, however only in 13.2% based on ESCAT.

Conclusions: Our study identifies known and novel prognostic and predictive biomarker candidates in patients with mCRPC undergoing first-line treatment with enzalutamide or abiraterone acetate. It further provides real-world evidence of the significant potential of genomic profiling of ctDNA to inform treatment in this setting. Clinical trials are warranted to advance the implementation of ctDNA-based biomarkers into clinical practice.

Keywords: Biomarker; Circulating cell-free DNA (cfDNA); Circulating tumor DNA (ctDNA); Clinical actionability; Clonal hematopoiesis; Liquid biopsy; Metastatic castration-resistant prostate cancer (mCRPC); Predictive; Prognostic; Resistance.

Fig. 1

Overview of the most frequently altered genes in baseline samples from mCRPC patients. Genes with somatic or germline alterations in at least 3 patients are displayed. Only pathogenic and likely pathogenic SNVs, SNVs annotated as moderate or high impact, structural variants, amplifications, and heterozygous and homozygous deletions are shown. Patients are ordered according to months to PSA progression (top barplot, red line indicates cut-off for primary resistance defined as treatment failure within the first three months). ctDNA fraction as determined by ichorCNA is shown in the bottom barplot. *Indicates patients that were censored (see also Table 1). (AMP, amplification; HET-DEL, heterozygous deletion; HOM-DEL, homozygous deletion; MSI, microsatellite instability; SNV, small nucleotide variant; SV, structural variant)

Fig. 2

Genomic correlates of clinical outcomes. a Univariate and multivariate Cox regression using PSA PFS and OS as endpoints. Multivariate analyses corrected for ctDNA fraction. b-d Kaplan Meier plots of patients with alterations in PTEN, cell cycle regulators (CCND1, CDKN1B, CDKN2A, RB1), or chromatin modulators (ARID1A, CHD1), compared to patients without alterations, using PSA PFS and OS as endpoints. e-g Kaplan-Meier plots for same genes as in b-d, however employing publicly available data from Annala et al. [48]. P-values in Kaplan-Meier plots based on log-rank test. Only pathogenic and likely pathogenic SNVs, SNVs annotated as high impact variants, amplifications, and homozygous deletions were included (i.e., structural variants and heterozygous deletions were excluded)

Fig. 3

Genomic correlates of intrinsic resistance to second-generation anti-androgens. a Frequency of alterations in patients with primary resistance to first-line enzalutamide or abiraterone acetate compared to patients without. Primary resistance was defined as treatment failure by 3 months, and only genes with alterations in at least two patients were considered. (Fisher’s exact test, *p-value < 0.05) b Comparison of the mutational frequencies of alterations in patients with primary resistance to enzalutamide or abiraterone acetate relative to patients that responded to treatment. The difference in relative frequency is shown on the x-axis and the -log10(p-value) (Fisher’s exact test) is shown on the y-axis. Genes that were significantly enriched for in non-responders are shown in red and those enriched for in responders in blue (p-value < 0.05). c Frequency of alterations in patients with primary resistance for genes in (a) based on the Annala et al. cohort [48] d Analysis as in (b), however based on the Annala et al. dataset

Fig. 4

Emerging genomic alterations during first-line treatment with second generation anti-androgens. a Oncoplot of variants emerging at progression (n = 18). b Overview of variants in matched baseline and progression samples from four patients that acquired AR resistance mutations at progression. VAFs for the corresponding variants, as well as changes in PSA and ctDNA% from baseline to progression are shown. c Patient-specific ddPCR assays showing the longitudinal dynamics of AR resistance mutations. VAF based on ddPCR is shown, as well as PSA changes from baseline to progression. Open circles represent time points where the variant was not detected based on ddPCR. Shaded region indicates time from initial PSA progression to treatment discontinuation. (AMP, amplification; HET-DEL, heterozygous deletion; HOM-DEL, homozygous deletion; SNV, small nucleotide variant; SV, structural variant). *SNV present at baseline

Fig. 5

Clinically actionable alterations. a Proportion of clinically actionable alterations based on the highest-level alterations identified for each patient, annotated based on OncoKB and ESCAT (the different levels of clinically actionable alterations according to OncoKB are shown adapted from https://www.oncokb.org/, and for ESCAT from Mateo et al. [40] and Mosele et al. [41]. Only relevant tiers are noted herein). b Overview of clinically actionable alterations detected in the cohort and as annotated by OncoKB and ESCAT. (AMP, amplification; HET-DEL, heterozygous deletion; HOM-DEL, homozygous deletion; MMR, mismatch repair; SNV, small nucleotide variant; SV, structural variant)